Even with your settings, it is possible to discern the existence of a scattered
field, when no scatterer is present.
I think the problem isn't numerical dispersion per se, but rather the discrete
nature of your sources. I think if you use line sources instead of individual,
discrete point sources, this should disappear.
One thing I didn't understand were the cy and cx parameters. Also, in the
first part of your original code you had 4 pixel shift which doesn't appear in
the second part. Was there something different?
Best,
Matt
On Wed, 30 Jan 2008, Andreas Unger wrote:
> Hi matt,
>
> This seems to be a problem of numerical dispersion introduced by the
> finite grid. When I change the parameters to
>
> (define-param freq 5);frequency
> (define-param fcut 5);cutoff
> (define-param fw 0.1);frequency width at turning on the source
>
> (define Lpml 0.5) ;thickness of the pml
> (define L 2 ) ;total size of the space
> (define tfSize 0.9);edge length of total field area
>
> and
>
> (set! resolution 250);Set the resolution
>
> then the planewave looks nice again. Actually this TFSF works only fine
> for relativly small TF areas. The reason is that it requires very
> precise timing for turning on the sources and annihilate the outgoing
> waves. In the Amplitude functions along the boundary the analytic
> wavevectors for the given frequency are used but in reality the
> wavevectors are a bit different due to numerical dispersion. The bigger
> the TF area gets compared to your wavlength the stronger you will see
> this! Mybe this effect can be compensated by using slightly different
> wavevectors but I have not tried.
>
> Greetings,
> Andreas
>
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
